Guides

The practical half.

Agent work fails on workflow long before it fails on the model: two sessions in one checkout, no exit condition, a diff nobody read. These are the fixes, with the git commands written out.

Guide

A git worktree workflow for coding agents

Why one worktree per agent is the difference between parallel work and a merge conflict, and how to set it up by hand in four git commands — with or without a workspace in front of it.

Read

Guide

How to run several coding agents in parallel without chaos

Splitting work across agents only pays off if the tasks are independent and the review keeps up. A practical pattern for parallel agent work, including the three ways it usually breaks.

Read

Guide

Reviewing an agent’s diff before you merge it

An agent’s work is a pull request without a person behind it. What to look for first, what agents get wrong most often, and why the review has to happen before the merge rather than after.

Read

Guide

Claude Code or Codex: picking the agent per task

Both are terminal agents and both are good. The useful question is not which one wins but which one to hand this task — and how to run them side by side without a merge fight.

Read

Guide

Running coding agents on Windows

Terminal agents assume a Unix shell. What breaks on Windows — PTY, line endings, paths, worktrees — and what a working setup looks like.

Read

Guide

What is an agent workspace?

A definition, the four parts that make one work — tasks, checkouts, terminals, review — and how it differs from an IDE with an agent in it or a terminal with AI in it.

Read